2

Google has made an announce about its new Willow quantum chip.

Here is their roadmap :enter image description here

which claims to have reached efficient error correction with 100 qubits.

On the other hand, I was aware of IBM's roadmap : enter image description here

which plans error correction enabled QPUs by 2029 (in 4 years), with 200 qubits.

I am quite surprised that Google has already reached this milestone comparatively to IBM, as they use the same qubit technology, superconducting circuits (perhaps with some minor differences) and that IBM target, 200 qubits, is in the same order of magnitude as Google.

As far as I understand, IBM develops its QPUs with multipurpose in mind, i.e. they do not restrict their usage to some kind of problems like e.g., D-Wave quantum Annealers for QUBO.

Maybe, Google also restricts its QPU development to a particular type of problem ? AI related ? which would explain why Google was faster to develop Error Correction enabled QPUs ?

glS
  • 27,510
  • 7
  • 37
  • 125
deb2014
  • 431
  • 2
  • 6

1 Answers1

5

Both IBM and Google develop universal, superconducting quantum computers. Both don’t “restrict its QPU development to a particular type of problem”, as an answer to your question (not that I aware of, at least).

As for IBM’s roadmap, I think they mean that they plan to deliver a quantum system named “Starling” with 200 logical qubits by 2029. Google encoded one (relatively noisy) logical qubit using ~100 physical qubits. My honest opinion - I will be very surprised if indeed anyone will be able to deliver high-quality 200 superconducting-based logical qubits by 2029.

As for the differences between IBM’s and Google’s QPUs - I think the most evident difference is the topologies. Google constantly present better qubit-connectivity while manifesting grid topology (which is suitable for the “surface” quantum error correction code, as they demonstrated with Willow):

enter image description here

IBM constantly present QPUs manifesting the “heavy-hex” topology - with poor qubit-connectivity and not suitable for the surface code:

enter image description here

Fabricating high-fidelity qubits with high connectivity compared to low connectivity is much harder due to crosstalks and inevitable undesired interactions between the connected components.

BTW, until recently IBM used fixed-frequency qubits and simple capacitors as couplers, while Google fabricated flux-tunable qubits and couplers already in their 2019 supremacy experiment. Right now both IBM and Google use flux-tunable Transmons as qubits and couplers, as far as I know.

Ohad
  • 1,899
  • 3
  • 15